Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x21a0634f9b0>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x21a0640a160>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.2.1
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # DONE: Implement Function
    inputs_real = tf.placeholder(tf.float32, (None, image_width, image_height, image_channels), name='input_real') 
    inputs_z = tf.placeholder(tf.float32, (None, z_dim), name='input_z')
    lr = tf.placeholder(tf.float32, name='learning_rate')
    
    return inputs_real, inputs_z, lr

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

ERROR:tensorflow:==================================
Object was never used (type <class 'tensorflow.python.framework.ops.Operation'>):
<tf.Operation 'assert_rank_2/Assert/Assert' type=Assert>
If you want to mark it as used call its "mark_used()" method.
It was originally created here:
['File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\runpy.py", line 193, in _run_module_as_main\n    "__main__", mod_spec)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\runpy.py", line 85, in _run_code\n    exec(code, run_globals)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel_launcher.py", line 16, in <module>\n    app.launch_new_instance()', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\traitlets\\config\\application.py", line 658, in launch_instance\n    app.start()', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\kernelapp.py", line 477, in start\n    ioloop.IOLoop.instance().start()', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\zmq\\eventloop\\ioloop.py", line 177, in start\n    super(ZMQIOLoop, self).start()', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tornado\\ioloop.py", line 888, in start\n    handler_func(fd_obj, events)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tornado\\stack_context.py", line 277, in null_wrapper\n    return fn(*args, **kwargs)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 440, in _handle_events\n    self._handle_recv()', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 472, in _handle_recv\n    self._run_callback(callback, msg)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\zmq\\eventloop\\zmqstream.py", line 414, in _run_callback\n    callback(*args, **kwargs)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tornado\\stack_context.py", line 277, in null_wrapper\n    return fn(*args, **kwargs)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\kernelbase.py", line 283, in dispatcher\n    return self.dispatch_shell(stream, msg)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\kernelbase.py", line 235, in dispatch_shell\n    handler(stream, idents, msg)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\kernelbase.py", line 399, in execute_request\n    user_expressions, allow_stdin)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\ipkernel.py", line 196, in do_execute\n    res = shell.run_cell(code, store_history=store_history, silent=silent)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\ipykernel\\zmqshell.py", line 533, in run_cell\n    return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2698, in run_cell\n    interactivity=interactivity, compiler=compiler, result=result)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2808, in run_ast_nodes\n    if self.run_code(code, result):', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\IPython\\core\\interactiveshell.py", line 2862, in run_code\n    exec(code_obj, self.user_global_ns, self.user_ns)', 'File "<ipython-input-6-6f2affc47cb0>", line 22, in <module>\n    tests.test_model_inputs(model_inputs)', 'File "C:\\Users\\peter\\Projects\\DeepLearning\\deep-learning\\face_generation\\problem_unittests.py", line 12, in func_wrapper\n    result = func(*args)', 'File "C:\\Users\\peter\\Projects\\DeepLearning\\deep-learning\\face_generation\\problem_unittests.py", line 68, in test_model_inputs\n    _check_input(learn_rate, [], \'Learning Rate\')', 'File "C:\\Users\\peter\\Projects\\DeepLearning\\deep-learning\\face_generation\\problem_unittests.py", line 34, in _check_input\n    _assert_tensor_shape(tensor, shape, \'Real Input\')', 'File "C:\\Users\\peter\\Projects\\DeepLearning\\deep-learning\\face_generation\\problem_unittests.py", line 20, in _assert_tensor_shape\n    assert tf.assert_rank(tensor, len(shape), message=\'{} has wrong rank\'.format(display_name))', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tensorflow\\python\\ops\\check_ops.py", line 617, in assert_rank\n    dynamic_condition, data, summarize)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tensorflow\\python\\ops\\check_ops.py", line 571, in _assert_rank_condition\n    return control_flow_ops.Assert(condition, data, summarize=summarize)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 170, in wrapped\n    return _add_should_use_warning(fn(*args, **kwargs))', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 139, in _add_should_use_warning\n    wrapped = TFShouldUseWarningWrapper(x)', 'File "c:\\users\\peter\\anaconda3\\envs\\tflearn-gpu\\lib\\site-packages\\tensorflow\\python\\util\\tf_should_use.py", line 96, in __init__\n    stack = [s.strip() for s in traceback.format_stack()]']
==================================
Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

# conv2d uses xavier initializer (as the None is transformed in a new glorot_uniform_initializer after line 220 in https://github.com/tensorflow/tensorflow/blob/r1.2/tensorflow/python/ops/variable_scope.py)
dnconv = 64
dropout_rate = 0.5
def discriminator(images, reuse=False, alpha=0.01):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    # DONE: Implement Function
    with tf.variable_scope('discriminator', reuse=reuse):
        # Input layer is 28 x 28 x out_channel_dim
        #print(images.get_shape().as_list())
        
        # conv to 14, 14
        x = tf.layers.conv2d(images, dnconv*4, 5, strides=2, padding='same')
        x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # conv to 7, 7
        x = tf.layers.conv2d(x, dnconv*4, 5, strides=2, padding='same', use_bias=False)
        x = tf.layers.batch_normalization(x, training=True)
        x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # conv to 4, 4
        x = tf.layers.conv2d(x, dnconv*8, 5, strides=2, padding='same', use_bias=False)
        x = tf.layers.batch_normalization(x, training=True)
        #x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # flatten
        x = tf.reshape(x, (-1, 4*4*dnconv*8))
        #print(x.get_shape().as_list())
        
        # logits and output
        logits = tf.layers.dense(x, 1)
        out = tf.sigmoid(logits)
        
        
        return out, logits
    
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

gnconv = 64
dropout_rate = 0.5
def generator(z, out_channel_dim, is_train=True, alpha=0.02): # I put alpha parameter to this function call
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    # DONE: Implement Function
    
    # auxiliar variables
    reuse = not is_train
    
    # scope for generator
    with tf.variable_scope('generator', reuse=reuse):
        # First fully connected layer
        x = tf.layers.dense(z, 4*4*gnconv*8)
        #print(x.get_shape().as_list())
        
        # batch normalization to 4x4
        x = tf.reshape(x, (-1, 4, 4, gnconv*8))
        x = tf.layers.batch_normalization(x, training=is_train)
        x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # conv to 2x2
        #x = tf.layers.conv2d_transpose(x, gnconv*8, 5, strides=2, padding='same')
        #x = tf.layers.batch_normalization(x, training=is_train)
        #x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        #x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # conv to 4x4
        #x = tf.layers.conv2d_transpose(x, gnconv*8, 5, strides=2, padding='same')
        #x = tf.layers.batch_normalization(x, training=is_train)
        #x = tf.layers.dropout(inputs=x, rate=dropout_rate, training=True)
        #x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())

        # upsample to 7x7
        x = tf.image.resize_nearest_neighbor(x, (7,7))
        #print(x.get_shape().as_list())

        # conv to 7x7
        x = tf.layers.conv2d(x, gnconv*8, 5, padding='same')
        x = tf.layers.batch_normalization(x, training=is_train)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # upsample to 14x14
        x = tf.image.resize_nearest_neighbor(x, (14,14))
        #print(x.get_shape().as_list())

        # conv to 14x14
        x = tf.layers.conv2d(x, gnconv*6, 5, padding='same')
        x = tf.layers.batch_normalization(x, training=is_train)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # upsample to 21x21
        x = tf.image.resize_nearest_neighbor(x, (21,21))
        #print(x.get_shape().as_list())

        # conv to 21x21
        x = tf.layers.conv2d(x, gnconv*4, 5, padding='same')
        x = tf.layers.batch_normalization(x, training=is_train)
        x = tf.maximum(alpha * x, x)
        #print(x.get_shape().as_list())
        
        # upsample to 28x28
        x = tf.image.resize_nearest_neighbor(x, (28,28))
        #print(x.get_shape().as_list())

        # Output layer, 28 x 28 x out_channel_dim (3 or 1, depending on the task - faces or minst)
        logits = tf.layers.conv2d(x, out_channel_dim, 5, padding='same')
        #print(logits.get_shape().as_list())
        
        out = tf.tanh(logits)
        
        return out

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

smooth = 0.1

def model_loss(input_real, input_z, out_channel_dim, alpha=0.02): # I put alpha parameter to this function call
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    # DONE: Implement Function
    
    # Just for reference:
    #   discriminator(images, reuse=False, alpha=0.01)
    #   generator(z, out_channel_dim, is_train=True, alpha=0.02): # I put alpha parameter to this function call
    
    g_out = generator(input_z, out_channel_dim, alpha=alpha)
    d_out_real, d_logits_real = discriminator(input_real, alpha=alpha)
    d_out_fake, d_logits_fake = discriminator(g_out, reuse=True, alpha=alpha)

    d_loss_real = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
                                                        logits=d_logits_real,
                                                        labels=tf.ones_like(d_out_real) * (1. - smooth)))
    d_loss_fake = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
                                                        logits=d_logits_fake,
                                                        labels=tf.zeros_like(d_out_fake)))
    d_loss = d_loss_real + d_loss_fake
    
    g_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
                                                        logits=d_logits_fake,
                                                        labels=tf.ones_like(d_out_fake)))

    return d_loss, g_loss

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # DONE: Implement Function
    
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

from datetime import datetime

print_every = 10
show_every = 100
n_images = 16
def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    # DONE: Build Model
    minput_real, minput_z, mlearning_rate = model_inputs(*data_shape[1:], z_dim)
    d_loss, g_loss = model_loss(minput_real, minput_z, data_shape[-1], alpha=0.02)
    d_opt, g_opt = model_opt(d_loss, g_loss, learning_rate, beta1)
    have_i_already_reset_the_default_graph = True

    # Auxiliar variables
    steps = 0
    out_channel_dim = 3 if data_image_mode == 'RGB' else 1
    sample_z = np.random.uniform(-1, 1, size=(n_images, z_dim))

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):    
            for batch_images in get_batches(batch_size):
                # DONE: Train Model
                steps += 1
                
                # rescale from [-.5, .5] to [-1., 1.]
                batch_images = batch_images * 2

                # Sample random noise for G
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))

                # Run optimizers
                #_ = sess.run(net.d_opt, feed_dict={net.input_real: batch_images, net.input_z: batch_z, net.learning_rate : learning_rate})
                #_ = sess.run(net.g_opt, feed_dict={net.input_z: batch_z, net.learning_rate : learning_rate, net.input_real: batch_images})
                _ = sess.run(d_opt, feed_dict={minput_real: batch_images, minput_z: batch_z, mlearning_rate: learning_rate})
                _ = sess.run(g_opt, feed_dict={minput_z: batch_z, mlearning_rate: learning_rate, minput_real: batch_images})

                if steps % print_every == 0:
                    # At the end of each epoch, get the losses and print them out
                    #train_loss_d = net.d_loss.eval({net.input_z: batch_z, net.input_real: batch_images})
                    #train_loss_g = net.g_loss.eval({net.input_z: batch_z})
                    train_loss_d = d_loss.eval({minput_z: batch_z, minput_real: batch_images})
                    train_loss_g = g_loss.eval({minput_z: batch_z})

                    print("Timestamp {} - ".format(datetime.now().strftime('%Y%m%d-%H-%M-%S.%f')[:-3]),
                          "Epoch {}/{} - ".format(epoch_i+1, epochs),
                          "Step {} - ".format(steps),
                          "D Loss: {:.4f} -".format(train_loss_d),
                          "G Loss: {:.4f}".format(train_loss_g))

                if steps % show_every == 0:
                    show_generator_output(sess, n_images, minput_z, out_channel_dim, data_image_mode)

            # At the end of each epoch, get the losses and print them out
            if not (steps % show_every == 0):
                train_loss_d = d_loss.eval({minput_z: batch_z, minput_real: batch_images})
                train_loss_g = g_loss.eval({minput_z: batch_z})
                print("Timestamp {} - ".format(datetime.now().strftime('%Y%m%d-%H-%M-%S.%f')[:-3]),
                      "Epoch {}/{} - ".format(epoch_i+1, epochs),
                      "Step {} - ".format(steps),
                      "D Loss: {:.4f} -".format(train_loss_d),
                      "G Loss: {:.4f}".format(train_loss_g))

                show_generator_output(sess, n_images, minput_z, out_channel_dim, data_image_mode)

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 8
z_dim = 100
learning_rate = 0.0002
beta1 = 0.5

tf.reset_default_graph()

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Timestamp 20170806-10-23-27.713 -  Epoch 1/2 -  Step 10 -  D Loss: 0.4946 - G Loss: 2.1428
Timestamp 20170806-10-23-39.550 -  Epoch 1/2 -  Step 20 -  D Loss: 1.9922 - G Loss: 1.7634
Timestamp 20170806-10-23-51.376 -  Epoch 1/2 -  Step 30 -  D Loss: 5.0702 - G Loss: 0.0272
Timestamp 20170806-10-24-03.203 -  Epoch 1/2 -  Step 40 -  D Loss: 1.8834 - G Loss: 2.7380
Timestamp 20170806-10-24-15.025 -  Epoch 1/2 -  Step 50 -  D Loss: 2.3946 - G Loss: 1.0202
Timestamp 20170806-10-24-26.856 -  Epoch 1/2 -  Step 60 -  D Loss: 1.1660 - G Loss: 6.4386
Timestamp 20170806-10-24-38.694 -  Epoch 1/2 -  Step 70 -  D Loss: 1.6963 - G Loss: 0.4441
Timestamp 20170806-10-24-50.530 -  Epoch 1/2 -  Step 80 -  D Loss: 2.1858 - G Loss: 0.8111
Timestamp 20170806-10-25-02.348 -  Epoch 1/2 -  Step 90 -  D Loss: 0.7884 - G Loss: 3.0378
Timestamp 20170806-10-25-14.171 -  Epoch 1/2 -  Step 100 -  D Loss: 0.7693 - G Loss: 3.2450

Timestamp 20170806-10-25-26.762 -  Epoch 1/2 -  Step 110 -  D Loss: 3.0906 - G Loss: 0.6487
Timestamp 20170806-10-25-38.583 -  Epoch 1/2 -  Step 120 -  D Loss: 1.1779 - G Loss: 1.4992
Timestamp 20170806-10-25-50.419 -  Epoch 1/2 -  Step 130 -  D Loss: 1.2211 - G Loss: 5.4736
Timestamp 20170806-10-26-02.241 -  Epoch 1/2 -  Step 140 -  D Loss: 1.3144 - G Loss: 2.5306
Timestamp 20170806-10-26-14.080 -  Epoch 1/2 -  Step 150 -  D Loss: 1.5099 - G Loss: 0.8154
Timestamp 20170806-10-26-25.910 -  Epoch 1/2 -  Step 160 -  D Loss: 2.3690 - G Loss: 1.1603
Timestamp 20170806-10-26-37.733 -  Epoch 1/2 -  Step 170 -  D Loss: 1.2669 - G Loss: 2.1021
Timestamp 20170806-10-26-49.565 -  Epoch 1/2 -  Step 180 -  D Loss: 1.6712 - G Loss: 1.3893
Timestamp 20170806-10-27-01.377 -  Epoch 1/2 -  Step 190 -  D Loss: 2.9108 - G Loss: 1.0272
Timestamp 20170806-10-27-13.218 -  Epoch 1/2 -  Step 200 -  D Loss: 2.8826 - G Loss: 0.5386

Timestamp 20170806-10-27-25.663 -  Epoch 1/2 -  Step 210 -  D Loss: 1.7954 - G Loss: 1.0714
Timestamp 20170806-10-27-37.501 -  Epoch 1/2 -  Step 220 -  D Loss: 1.2007 - G Loss: 1.1677
Timestamp 20170806-10-27-49.324 -  Epoch 1/2 -  Step 230 -  D Loss: 1.3387 - G Loss: 1.9886
Timestamp 20170806-10-28-01.149 -  Epoch 1/2 -  Step 240 -  D Loss: 1.3441 - G Loss: 1.3638
Timestamp 20170806-10-28-12.974 -  Epoch 1/2 -  Step 250 -  D Loss: 1.5585 - G Loss: 0.6861
Timestamp 20170806-10-28-24.807 -  Epoch 1/2 -  Step 260 -  D Loss: 1.3818 - G Loss: 0.7226
Timestamp 20170806-10-28-36.644 -  Epoch 1/2 -  Step 270 -  D Loss: 1.6154 - G Loss: 1.5620
Timestamp 20170806-10-28-48.458 -  Epoch 1/2 -  Step 280 -  D Loss: 1.4483 - G Loss: 1.2354
Timestamp 20170806-10-29-00.276 -  Epoch 1/2 -  Step 290 -  D Loss: 1.3838 - G Loss: 1.0379
Timestamp 20170806-10-29-12.090 -  Epoch 1/2 -  Step 300 -  D Loss: 0.9632 - G Loss: 1.9213

Timestamp 20170806-10-29-24.550 -  Epoch 1/2 -  Step 310 -  D Loss: 1.2788 - G Loss: 1.1211
Timestamp 20170806-10-29-36.384 -  Epoch 1/2 -  Step 320 -  D Loss: 0.9810 - G Loss: 1.6539
Timestamp 20170806-10-29-48.205 -  Epoch 1/2 -  Step 330 -  D Loss: 1.3301 - G Loss: 0.8778
Timestamp 20170806-10-30-00.023 -  Epoch 1/2 -  Step 340 -  D Loss: 1.4658 - G Loss: 1.4267
Timestamp 20170806-10-30-11.830 -  Epoch 1/2 -  Step 350 -  D Loss: 1.8681 - G Loss: 1.6740
Timestamp 20170806-10-30-23.640 -  Epoch 1/2 -  Step 360 -  D Loss: 1.7254 - G Loss: 0.9603
Timestamp 20170806-10-30-35.467 -  Epoch 1/2 -  Step 370 -  D Loss: 1.6644 - G Loss: 1.2092
Timestamp 20170806-10-30-47.300 -  Epoch 1/2 -  Step 380 -  D Loss: 0.9235 - G Loss: 1.5786
Timestamp 20170806-10-30-59.128 -  Epoch 1/2 -  Step 390 -  D Loss: 1.1464 - G Loss: 1.3533
Timestamp 20170806-10-31-10.953 -  Epoch 1/2 -  Step 400 -  D Loss: 1.5339 - G Loss: 0.7561

Timestamp 20170806-10-31-23.402 -  Epoch 1/2 -  Step 410 -  D Loss: 1.6426 - G Loss: 1.0696
Timestamp 20170806-10-31-35.216 -  Epoch 1/2 -  Step 420 -  D Loss: 1.2460 - G Loss: 1.2256
Timestamp 20170806-10-31-47.023 -  Epoch 1/2 -  Step 430 -  D Loss: 1.4338 - G Loss: 1.0133
Timestamp 20170806-10-31-58.848 -  Epoch 1/2 -  Step 440 -  D Loss: 1.3872 - G Loss: 0.8139
Timestamp 20170806-10-32-10.667 -  Epoch 1/2 -  Step 450 -  D Loss: 1.2630 - G Loss: 1.8291
Timestamp 20170806-10-32-22.484 -  Epoch 1/2 -  Step 460 -  D Loss: 1.3318 - G Loss: 0.6638
Timestamp 20170806-10-32-34.308 -  Epoch 1/2 -  Step 470 -  D Loss: 1.3119 - G Loss: 0.6439
Timestamp 20170806-10-32-46.133 -  Epoch 1/2 -  Step 480 -  D Loss: 1.6917 - G Loss: 0.7436
Timestamp 20170806-10-32-57.950 -  Epoch 1/2 -  Step 490 -  D Loss: 1.8414 - G Loss: 0.8627
Timestamp 20170806-10-33-09.772 -  Epoch 1/2 -  Step 500 -  D Loss: 1.7589 - G Loss: 1.1794

Timestamp 20170806-10-33-22.244 -  Epoch 1/2 -  Step 510 -  D Loss: 1.3109 - G Loss: 0.5671
Timestamp 20170806-10-33-34.060 -  Epoch 1/2 -  Step 520 -  D Loss: 1.1844 - G Loss: 1.0964
Timestamp 20170806-10-33-45.871 -  Epoch 1/2 -  Step 530 -  D Loss: 1.7141 - G Loss: 0.8166
Timestamp 20170806-10-33-57.712 -  Epoch 1/2 -  Step 540 -  D Loss: 1.6052 - G Loss: 0.9179
Timestamp 20170806-10-34-09.540 -  Epoch 1/2 -  Step 550 -  D Loss: 1.4786 - G Loss: 0.6799
Timestamp 20170806-10-34-21.350 -  Epoch 1/2 -  Step 560 -  D Loss: 1.3231 - G Loss: 1.2949
Timestamp 20170806-10-34-33.172 -  Epoch 1/2 -  Step 570 -  D Loss: 1.6371 - G Loss: 0.9100
Timestamp 20170806-10-34-44.975 -  Epoch 1/2 -  Step 580 -  D Loss: 1.2354 - G Loss: 1.1922
Timestamp 20170806-10-34-56.801 -  Epoch 1/2 -  Step 590 -  D Loss: 1.1620 - G Loss: 0.8592
Timestamp 20170806-10-35-08.604 -  Epoch 1/2 -  Step 600 -  D Loss: 1.6745 - G Loss: 0.7188

Timestamp 20170806-10-35-21.059 -  Epoch 1/2 -  Step 610 -  D Loss: 1.2532 - G Loss: 1.0650
Timestamp 20170806-10-35-32.880 -  Epoch 1/2 -  Step 620 -  D Loss: 1.3126 - G Loss: 1.0756
Timestamp 20170806-10-35-44.694 -  Epoch 1/2 -  Step 630 -  D Loss: 1.4675 - G Loss: 1.1660
Timestamp 20170806-10-35-56.514 -  Epoch 1/2 -  Step 640 -  D Loss: 1.2891 - G Loss: 0.9819
Timestamp 20170806-10-36-08.336 -  Epoch 1/2 -  Step 650 -  D Loss: 1.4598 - G Loss: 0.8097
Timestamp 20170806-10-36-20.158 -  Epoch 1/2 -  Step 660 -  D Loss: 1.5426 - G Loss: 1.0653
Timestamp 20170806-10-36-31.985 -  Epoch 1/2 -  Step 670 -  D Loss: 1.2679 - G Loss: 1.0072
Timestamp 20170806-10-36-43.808 -  Epoch 1/2 -  Step 680 -  D Loss: 1.3361 - G Loss: 1.0939
Timestamp 20170806-10-36-55.614 -  Epoch 1/2 -  Step 690 -  D Loss: 1.0014 - G Loss: 1.0909
Timestamp 20170806-10-37-07.436 -  Epoch 1/2 -  Step 700 -  D Loss: 1.4344 - G Loss: 1.1159

Timestamp 20170806-10-37-19.908 -  Epoch 1/2 -  Step 710 -  D Loss: 1.1071 - G Loss: 1.3559
Timestamp 20170806-10-37-31.727 -  Epoch 1/2 -  Step 720 -  D Loss: 1.2324 - G Loss: 1.0698
Timestamp 20170806-10-37-43.555 -  Epoch 1/2 -  Step 730 -  D Loss: 1.2397 - G Loss: 1.3712
Timestamp 20170806-10-37-55.381 -  Epoch 1/2 -  Step 740 -  D Loss: 1.2716 - G Loss: 1.2853
Timestamp 20170806-10-38-07.197 -  Epoch 1/2 -  Step 750 -  D Loss: 1.4365 - G Loss: 0.6773
Timestamp 20170806-10-38-19.011 -  Epoch 1/2 -  Step 760 -  D Loss: 1.4794 - G Loss: 0.7840
Timestamp 20170806-10-38-30.824 -  Epoch 1/2 -  Step 770 -  D Loss: 1.2564 - G Loss: 1.0047
Timestamp 20170806-10-38-42.640 -  Epoch 1/2 -  Step 780 -  D Loss: 1.3363 - G Loss: 0.9613
Timestamp 20170806-10-38-54.455 -  Epoch 1/2 -  Step 790 -  D Loss: 0.9653 - G Loss: 0.9729
Timestamp 20170806-10-39-06.264 -  Epoch 1/2 -  Step 800 -  D Loss: 1.7702 - G Loss: 0.8116

Timestamp 20170806-10-39-18.726 -  Epoch 1/2 -  Step 810 -  D Loss: 1.1893 - G Loss: 1.3364
Timestamp 20170806-10-39-30.554 -  Epoch 1/2 -  Step 820 -  D Loss: 1.0703 - G Loss: 0.7858
Timestamp 20170806-10-39-42.386 -  Epoch 1/2 -  Step 830 -  D Loss: 1.1172 - G Loss: 1.1571
Timestamp 20170806-10-39-54.184 -  Epoch 1/2 -  Step 840 -  D Loss: 0.9958 - G Loss: 1.2377
Timestamp 20170806-10-40-05.993 -  Epoch 1/2 -  Step 850 -  D Loss: 1.4729 - G Loss: 0.6245
Timestamp 20170806-10-40-17.832 -  Epoch 1/2 -  Step 860 -  D Loss: 1.0923 - G Loss: 0.9571
Timestamp 20170806-10-40-29.640 -  Epoch 1/2 -  Step 870 -  D Loss: 1.2538 - G Loss: 1.2865
Timestamp 20170806-10-40-41.469 -  Epoch 1/2 -  Step 880 -  D Loss: 0.9312 - G Loss: 0.9547
Timestamp 20170806-10-40-53.310 -  Epoch 1/2 -  Step 890 -  D Loss: 1.2358 - G Loss: 1.5055
Timestamp 20170806-10-41-05.155 -  Epoch 1/2 -  Step 900 -  D Loss: 1.3938 - G Loss: 1.1742

Timestamp 20170806-10-41-17.836 -  Epoch 1/2 -  Step 910 -  D Loss: 0.8351 - G Loss: 1.0277
Timestamp 20170806-10-41-29.690 -  Epoch 1/2 -  Step 920 -  D Loss: 0.9895 - G Loss: 0.9915
Timestamp 20170806-10-41-41.525 -  Epoch 1/2 -  Step 930 -  D Loss: 1.0083 - G Loss: 1.5915
Timestamp 20170806-10-41-53.379 -  Epoch 1/2 -  Step 940 -  D Loss: 1.3331 - G Loss: 0.8304
Timestamp 20170806-10-42-05.215 -  Epoch 1/2 -  Step 950 -  D Loss: 1.0908 - G Loss: 1.1948
Timestamp 20170806-10-42-17.055 -  Epoch 1/2 -  Step 960 -  D Loss: 1.2853 - G Loss: 0.9411
Timestamp 20170806-10-42-28.904 -  Epoch 1/2 -  Step 970 -  D Loss: 1.3128 - G Loss: 1.1966
Timestamp 20170806-10-42-40.749 -  Epoch 1/2 -  Step 980 -  D Loss: 1.3875 - G Loss: 1.2091
Timestamp 20170806-10-42-52.579 -  Epoch 1/2 -  Step 990 -  D Loss: 1.4521 - G Loss: 0.9095
Timestamp 20170806-10-43-04.401 -  Epoch 1/2 -  Step 1000 -  D Loss: 1.4762 - G Loss: 1.1141

batch_size = 16
z_dim = 100
learning_rate = 0.0002
beta1 = 0.5

tf.reset_default_graph()

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Timestamp 20170806-15-21-01.873 -  Epoch 1/2 -  Step 10 -  D Loss: 1.3090 - G Loss: 1.2414
Timestamp 20170806-15-21-18.283 -  Epoch 1/2 -  Step 20 -  D Loss: 1.8177 - G Loss: 0.6976
Timestamp 20170806-15-21-34.688 -  Epoch 1/2 -  Step 30 -  D Loss: 1.0075 - G Loss: 2.4195
Timestamp 20170806-15-21-51.089 -  Epoch 1/2 -  Step 40 -  D Loss: 1.1479 - G Loss: 1.4919
Timestamp 20170806-15-22-07.495 -  Epoch 1/2 -  Step 50 -  D Loss: 1.5415 - G Loss: 1.7543
Timestamp 20170806-15-22-23.928 -  Epoch 1/2 -  Step 60 -  D Loss: 0.6932 - G Loss: 3.5427
Timestamp 20170806-15-22-40.344 -  Epoch 1/2 -  Step 70 -  D Loss: 4.7297 - G Loss: 0.0600
Timestamp 20170806-15-22-56.738 -  Epoch 1/2 -  Step 80 -  D Loss: 1.7995 - G Loss: 1.6519
Timestamp 20170806-15-23-13.154 -  Epoch 1/2 -  Step 90 -  D Loss: 1.1145 - G Loss: 1.1342
Timestamp 20170806-15-23-29.551 -  Epoch 1/2 -  Step 100 -  D Loss: 0.8019 - G Loss: 2.1190

Timestamp 20170806-15-23-46.605 -  Epoch 1/2 -  Step 110 -  D Loss: 0.6483 - G Loss: 2.4192
Timestamp 20170806-15-24-03.000 -  Epoch 1/2 -  Step 120 -  D Loss: 1.4651 - G Loss: 1.0364
Timestamp 20170806-15-24-19.402 -  Epoch 1/2 -  Step 130 -  D Loss: 2.6827 - G Loss: 0.6201
Timestamp 20170806-15-24-35.827 -  Epoch 1/2 -  Step 140 -  D Loss: 2.3734 - G Loss: 1.1966
Timestamp 20170806-15-24-52.252 -  Epoch 1/2 -  Step 150 -  D Loss: 1.5409 - G Loss: 0.8949
Timestamp 20170806-15-25-08.671 -  Epoch 1/2 -  Step 160 -  D Loss: 2.5368 - G Loss: 0.7717
Timestamp 20170806-15-25-25.064 -  Epoch 1/2 -  Step 170 -  D Loss: 1.3127 - G Loss: 1.6137
Timestamp 20170806-15-25-41.471 -  Epoch 1/2 -  Step 180 -  D Loss: 1.5718 - G Loss: 1.0754
Timestamp 20170806-15-25-57.876 -  Epoch 1/2 -  Step 190 -  D Loss: 1.0825 - G Loss: 1.5391
Timestamp 20170806-15-26-14.296 -  Epoch 1/2 -  Step 200 -  D Loss: 1.3343 - G Loss: 1.3543

Timestamp 20170806-15-26-31.325 -  Epoch 1/2 -  Step 210 -  D Loss: 1.3657 - G Loss: 0.7775
Timestamp 20170806-15-26-47.750 -  Epoch 1/2 -  Step 220 -  D Loss: 1.7249 - G Loss: 0.8611
Timestamp 20170806-15-27-04.166 -  Epoch 1/2 -  Step 230 -  D Loss: 1.1352 - G Loss: 0.9989
Timestamp 20170806-15-27-20.573 -  Epoch 1/2 -  Step 240 -  D Loss: 1.6716 - G Loss: 1.3388
Timestamp 20170806-15-27-36.987 -  Epoch 1/2 -  Step 250 -  D Loss: 1.3549 - G Loss: 1.4990
Timestamp 20170806-15-27-53.398 -  Epoch 1/2 -  Step 260 -  D Loss: 1.5496 - G Loss: 0.7973
Timestamp 20170806-15-28-09.811 -  Epoch 1/2 -  Step 270 -  D Loss: 1.1064 - G Loss: 1.3357
Timestamp 20170806-15-28-26.217 -  Epoch 1/2 -  Step 280 -  D Loss: 1.1953 - G Loss: 0.8360
Timestamp 20170806-15-28-42.626 -  Epoch 1/2 -  Step 290 -  D Loss: 1.3890 - G Loss: 1.4177
Timestamp 20170806-15-28-59.030 -  Epoch 1/2 -  Step 300 -  D Loss: 1.2836 - G Loss: 1.3552

Timestamp 20170806-15-29-16.042 -  Epoch 1/2 -  Step 310 -  D Loss: 1.4008 - G Loss: 0.9672
Timestamp 20170806-15-29-32.449 -  Epoch 1/2 -  Step 320 -  D Loss: 0.8635 - G Loss: 1.6770
Timestamp 20170806-15-29-48.850 -  Epoch 1/2 -  Step 330 -  D Loss: 1.1921 - G Loss: 0.6360
Timestamp 20170806-15-30-05.231 -  Epoch 1/2 -  Step 340 -  D Loss: 0.9981 - G Loss: 1.5382
Timestamp 20170806-15-30-21.638 -  Epoch 1/2 -  Step 350 -  D Loss: 0.9204 - G Loss: 1.9675
Timestamp 20170806-15-30-38.031 -  Epoch 1/2 -  Step 360 -  D Loss: 1.4871 - G Loss: 1.3710
Timestamp 20170806-15-30-54.427 -  Epoch 1/2 -  Step 370 -  D Loss: 1.3198 - G Loss: 1.1309
Timestamp 20170806-15-31-10.828 -  Epoch 1/2 -  Step 380 -  D Loss: 0.8663 - G Loss: 1.5058
Timestamp 20170806-15-31-27.236 -  Epoch 1/2 -  Step 390 -  D Loss: 1.0862 - G Loss: 1.0735
Timestamp 20170806-15-31-43.657 -  Epoch 1/2 -  Step 400 -  D Loss: 1.5263 - G Loss: 2.6214

Timestamp 20170806-15-32-00.708 -  Epoch 1/2 -  Step 410 -  D Loss: 1.1131 - G Loss: 1.4967
Timestamp 20170806-15-32-17.110 -  Epoch 1/2 -  Step 420 -  D Loss: 1.9779 - G Loss: 0.7845
Timestamp 20170806-15-32-33.502 -  Epoch 1/2 -  Step 430 -  D Loss: 1.2244 - G Loss: 1.1163
Timestamp 20170806-15-32-49.900 -  Epoch 1/2 -  Step 440 -  D Loss: 1.2794 - G Loss: 1.4648
Timestamp 20170806-15-33-06.297 -  Epoch 1/2 -  Step 450 -  D Loss: 1.2701 - G Loss: 1.5669
Timestamp 20170806-15-33-22.713 -  Epoch 1/2 -  Step 460 -  D Loss: 1.3930 - G Loss: 0.6677
Timestamp 20170806-15-33-39.139 -  Epoch 1/2 -  Step 470 -  D Loss: 1.5468 - G Loss: 0.9789
Timestamp 20170806-15-33-55.561 -  Epoch 1/2 -  Step 480 -  D Loss: 1.5572 - G Loss: 0.9284
Timestamp 20170806-15-34-11.970 -  Epoch 1/2 -  Step 490 -  D Loss: 1.8506 - G Loss: 0.8442
Timestamp 20170806-15-34-28.382 -  Epoch 1/2 -  Step 500 -  D Loss: 1.4722 - G Loss: 1.0432

Timestamp 20170806-15-34-45.396 -  Epoch 1/2 -  Step 510 -  D Loss: 1.4360 - G Loss: 1.0467
Timestamp 20170806-15-35-01.798 -  Epoch 1/2 -  Step 520 -  D Loss: 1.0270 - G Loss: 1.3307
Timestamp 20170806-15-35-18.198 -  Epoch 1/2 -  Step 530 -  D Loss: 1.2593 - G Loss: 0.8764
Timestamp 20170806-15-35-34.590 -  Epoch 1/2 -  Step 540 -  D Loss: 1.3193 - G Loss: 1.2131
Timestamp 20170806-15-35-51.001 -  Epoch 1/2 -  Step 550 -  D Loss: 1.2009 - G Loss: 1.4451
Timestamp 20170806-15-36-07.397 -  Epoch 1/2 -  Step 560 -  D Loss: 1.3367 - G Loss: 0.9253
Timestamp 20170806-15-36-23.793 -  Epoch 1/2 -  Step 570 -  D Loss: 1.2794 - G Loss: 0.8036
Timestamp 20170806-15-36-40.182 -  Epoch 1/2 -  Step 580 -  D Loss: 0.9230 - G Loss: 1.5935
Timestamp 20170806-15-36-56.584 -  Epoch 1/2 -  Step 590 -  D Loss: 1.1712 - G Loss: 0.9888
Timestamp 20170806-15-37-12.977 -  Epoch 1/2 -  Step 600 -  D Loss: 1.0694 - G Loss: 0.8044

Timestamp 20170806-15-37-30.019 -  Epoch 1/2 -  Step 610 -  D Loss: 1.2986 - G Loss: 0.9396
Timestamp 20170806-15-37-46.431 -  Epoch 1/2 -  Step 620 -  D Loss: 0.9911 - G Loss: 1.2826
Timestamp 20170806-15-38-02.841 -  Epoch 1/2 -  Step 630 -  D Loss: 1.4266 - G Loss: 1.1482
Timestamp 20170806-15-38-19.234 -  Epoch 1/2 -  Step 640 -  D Loss: 1.1925 - G Loss: 0.6706
Timestamp 20170806-15-38-35.638 -  Epoch 1/2 -  Step 650 -  D Loss: 1.4823 - G Loss: 0.7950
Timestamp 20170806-15-38-52.042 -  Epoch 1/2 -  Step 660 -  D Loss: 1.1958 - G Loss: 1.0781
Timestamp 20170806-15-39-08.446 -  Epoch 1/2 -  Step 670 -  D Loss: 1.3415 - G Loss: 1.3041
Timestamp 20170806-15-39-24.850 -  Epoch 1/2 -  Step 680 -  D Loss: 1.3191 - G Loss: 1.0950
Timestamp 20170806-15-39-41.252 -  Epoch 1/2 -  Step 690 -  D Loss: 1.2116 - G Loss: 0.7195
Timestamp 20170806-15-39-57.658 -  Epoch 1/2 -  Step 700 -  D Loss: 1.2733 - G Loss: 1.1298

Timestamp 20170806-15-40-14.697 -  Epoch 1/2 -  Step 710 -  D Loss: 1.2906 - G Loss: 1.0336
Timestamp 20170806-15-40-31.104 -  Epoch 1/2 -  Step 720 -  D Loss: 1.0976 - G Loss: 0.8089
Timestamp 20170806-15-40-47.492 -  Epoch 1/2 -  Step 730 -  D Loss: 1.0531 - G Loss: 1.5428
Timestamp 20170806-15-41-03.899 -  Epoch 1/2 -  Step 740 -  D Loss: 1.3514 - G Loss: 0.7370
Timestamp 20170806-15-41-20.305 -  Epoch 1/2 -  Step 750 -  D Loss: 1.2040 - G Loss: 1.0151
Timestamp 20170806-15-41-36.691 -  Epoch 1/2 -  Step 760 -  D Loss: 1.2122 - G Loss: 1.1574
Timestamp 20170806-15-41-53.088 -  Epoch 1/2 -  Step 770 -  D Loss: 1.1407 - G Loss: 0.9774
Timestamp 20170806-15-42-09.484 -  Epoch 1/2 -  Step 780 -  D Loss: 1.1493 - G Loss: 1.1302
Timestamp 20170806-15-42-25.885 -  Epoch 1/2 -  Step 790 -  D Loss: 1.3655 - G Loss: 1.1568
Timestamp 20170806-15-42-42.294 -  Epoch 1/2 -  Step 800 -  D Loss: 1.4253 - G Loss: 1.2278

Timestamp 20170806-15-42-59.323 -  Epoch 1/2 -  Step 810 -  D Loss: 1.2934 - G Loss: 0.8237
Timestamp 20170806-15-43-15.729 -  Epoch 1/2 -  Step 820 -  D Loss: 1.2058 - G Loss: 0.9148
Timestamp 20170806-15-43-32.134 -  Epoch 1/2 -  Step 830 -  D Loss: 1.1873 - G Loss: 0.9021
Timestamp 20170806-15-43-48.542 -  Epoch 1/2 -  Step 840 -  D Loss: 1.1328 - G Loss: 1.1841
Timestamp 20170806-15-44-04.958 -  Epoch 1/2 -  Step 850 -  D Loss: 1.3153 - G Loss: 0.8495
Timestamp 20170806-15-44-21.367 -  Epoch 1/2 -  Step 860 -  D Loss: 1.1005 - G Loss: 1.1188
Timestamp 20170806-15-44-37.787 -  Epoch 1/2 -  Step 870 -  D Loss: 1.3392 - G Loss: 0.8239
Timestamp 20170806-15-44-54.174 -  Epoch 1/2 -  Step 880 -  D Loss: 1.3106 - G Loss: 1.1375
Timestamp 20170806-15-45-10.566 -  Epoch 1/2 -  Step 890 -  D Loss: 1.2776 - G Loss: 0.9591
Timestamp 20170806-15-45-26.952 -  Epoch 1/2 -  Step 900 -  D Loss: 1.2946 - G Loss: 1.0403

Timestamp 20170806-15-45-44.001 -  Epoch 1/2 -  Step 910 -  D Loss: 1.1015 - G Loss: 1.0806
Timestamp 20170806-15-46-00.401 -  Epoch 1/2 -  Step 920 -  D Loss: 1.1700 - G Loss: 0.9113
Timestamp 20170806-15-46-16.789 -  Epoch 1/2 -  Step 930 -  D Loss: 1.1949 - G Loss: 1.3002
Timestamp 20170806-15-46-33.195 -  Epoch 1/2 -  Step 940 -  D Loss: 1.0775 - G Loss: 0.9031
Timestamp 20170806-15-46-49.595 -  Epoch 1/2 -  Step 950 -  D Loss: 1.0384 - G Loss: 1.2036
Timestamp 20170806-15-47-05.987 -  Epoch 1/2 -  Step 960 -  D Loss: 1.2180 - G Loss: 1.0454
Timestamp 20170806-15-47-22.388 -  Epoch 1/2 -  Step 970 -  D Loss: 1.3301 - G Loss: 1.4003
Timestamp 20170806-15-47-38.782 -  Epoch 1/2 -  Step 980 -  D Loss: 1.2804 - G Loss: 1.0666
Timestamp 20170806-15-47-55.208 -  Epoch 1/2 -  Step 990 -  D Loss: 1.2386 - G Loss: 1.0048
Timestamp 20170806-15-48-11.623 -  Epoch 1/2 -  Step 1000 -  D Loss: 1.2002 - G Loss: 0.9387

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 8
z_dim = 100
learning_rate = 0.0002
beta1 = 0.5

tf.reset_default_graph()

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Timestamp 20170805-20-01-37.113 -  Epoch 1/1 -  Step 10 -  D Loss: 1.2554 - G Loss: 1.1563
Timestamp 20170805-20-02-00.672 -  Epoch 1/1 -  Step 20 -  D Loss: 0.9025 - G Loss: 2.2572
Timestamp 20170805-20-02-24.211 -  Epoch 1/1 -  Step 30 -  D Loss: 0.6287 - G Loss: 3.9705
Timestamp 20170805-20-02-47.781 -  Epoch 1/1 -  Step 40 -  D Loss: 0.8469 - G Loss: 5.2232
Timestamp 20170805-20-03-11.295 -  Epoch 1/1 -  Step 50 -  D Loss: 0.8443 - G Loss: 5.5423
Timestamp 20170805-20-03-34.855 -  Epoch 1/1 -  Step 60 -  D Loss: 0.7757 - G Loss: 1.1933
Timestamp 20170805-20-03-58.364 -  Epoch 1/1 -  Step 70 -  D Loss: 0.8121 - G Loss: 1.5073
Timestamp 20170805-20-04-21.908 -  Epoch 1/1 -  Step 80 -  D Loss: 0.5370 - G Loss: 3.8180
Timestamp 20170805-20-04-45.406 -  Epoch 1/1 -  Step 90 -  D Loss: 1.7351 - G Loss: 0.7592
Timestamp 20170805-20-05-08.931 -  Epoch 1/1 -  Step 100 -  D Loss: 0.6156 - G Loss: 4.2430

Timestamp 20170805-20-05-38.359 -  Epoch 1/1 -  Step 110 -  D Loss: 0.7066 - G Loss: 1.9523
Timestamp 20170805-20-06-01.867 -  Epoch 1/1 -  Step 120 -  D Loss: 1.9221 - G Loss: 0.4920
Timestamp 20170805-20-06-25.378 -  Epoch 1/1 -  Step 130 -  D Loss: 0.7178 - G Loss: 1.7770
Timestamp 20170805-20-06-48.891 -  Epoch 1/1 -  Step 140 -  D Loss: 1.7880 - G Loss: 1.2182
Timestamp 20170805-20-07-12.405 -  Epoch 1/1 -  Step 150 -  D Loss: 1.1891 - G Loss: 2.0380
Timestamp 20170805-20-07-35.918 -  Epoch 1/1 -  Step 160 -  D Loss: 0.5902 - G Loss: 3.0968
Timestamp 20170805-20-07-59.429 -  Epoch 1/1 -  Step 170 -  D Loss: 0.4528 - G Loss: 3.5085
Timestamp 20170805-20-08-22.938 -  Epoch 1/1 -  Step 180 -  D Loss: 1.1992 - G Loss: 1.5422
Timestamp 20170805-20-08-46.454 -  Epoch 1/1 -  Step 190 -  D Loss: 0.9280 - G Loss: 2.0220
Timestamp 20170805-20-09-09.968 -  Epoch 1/1 -  Step 200 -  D Loss: 2.9793 - G Loss: 0.5708

Timestamp 20170805-20-09-34.526 -  Epoch 1/1 -  Step 210 -  D Loss: 0.7018 - G Loss: 2.2072
Timestamp 20170805-20-09-58.034 -  Epoch 1/1 -  Step 220 -  D Loss: 0.4688 - G Loss: 3.6473
Timestamp 20170805-20-10-11.639 -  Epoch 1/1 -  Step 230 -  D Loss: 1.2144 - G Loss: 1.0949
Timestamp 20170805-20-10-23.572 -  Epoch 1/1 -  Step 240 -  D Loss: 0.7973 - G Loss: 2.4812
Timestamp 20170805-20-10-35.517 -  Epoch 1/1 -  Step 250 -  D Loss: 2.4051 - G Loss: 0.5535
Timestamp 20170805-20-10-47.435 -  Epoch 1/1 -  Step 260 -  D Loss: 1.0878 - G Loss: 1.4204
Timestamp 20170805-20-10-59.361 -  Epoch 1/1 -  Step 270 -  D Loss: 1.0537 - G Loss: 1.2385
Timestamp 20170805-20-11-11.310 -  Epoch 1/1 -  Step 280 -  D Loss: 1.3553 - G Loss: 3.0880
Timestamp 20170805-20-11-23.191 -  Epoch 1/1 -  Step 290 -  D Loss: 1.4201 - G Loss: 1.5190
Timestamp 20170805-20-11-35.096 -  Epoch 1/1 -  Step 300 -  D Loss: 1.0643 - G Loss: 0.9634

Timestamp 20170805-20-11-47.839 -  Epoch 1/1 -  Step 310 -  D Loss: 1.8670 - G Loss: 0.6469
Timestamp 20170805-20-11-59.748 -  Epoch 1/1 -  Step 320 -  D Loss: 1.4389 - G Loss: 1.0165
Timestamp 20170805-20-12-11.659 -  Epoch 1/1 -  Step 330 -  D Loss: 1.4505 - G Loss: 0.4686
Timestamp 20170805-20-12-23.596 -  Epoch 1/1 -  Step 340 -  D Loss: 0.9922 - G Loss: 2.1894
Timestamp 20170805-20-12-35.499 -  Epoch 1/1 -  Step 350 -  D Loss: 0.9329 - G Loss: 1.2147
Timestamp 20170805-20-12-47.411 -  Epoch 1/1 -  Step 360 -  D Loss: 1.3347 - G Loss: 1.1286
Timestamp 20170805-20-12-59.320 -  Epoch 1/1 -  Step 370 -  D Loss: 1.4106 - G Loss: 1.0294
Timestamp 20170805-20-13-11.225 -  Epoch 1/1 -  Step 380 -  D Loss: 0.6577 - G Loss: 1.5219
Timestamp 20170805-20-13-23.117 -  Epoch 1/1 -  Step 390 -  D Loss: 0.9870 - G Loss: 2.2122
Timestamp 20170805-20-13-35.030 -  Epoch 1/1 -  Step 400 -  D Loss: 1.0732 - G Loss: 2.3355

Timestamp 20170805-20-13-47.678 -  Epoch 1/1 -  Step 410 -  D Loss: 1.1106 - G Loss: 1.5170
Timestamp 20170805-20-13-59.589 -  Epoch 1/1 -  Step 420 -  D Loss: 0.9048 - G Loss: 2.1173
Timestamp 20170805-20-14-11.506 -  Epoch 1/1 -  Step 430 -  D Loss: 1.1099 - G Loss: 0.6701
Timestamp 20170805-20-14-23.399 -  Epoch 1/1 -  Step 440 -  D Loss: 1.3614 - G Loss: 3.0794
Timestamp 20170805-20-14-35.278 -  Epoch 1/1 -  Step 450 -  D Loss: 1.3076 - G Loss: 0.7063
Timestamp 20170805-20-14-47.181 -  Epoch 1/1 -  Step 460 -  D Loss: 1.0673 - G Loss: 1.5359
Timestamp 20170805-20-14-59.076 -  Epoch 1/1 -  Step 470 -  D Loss: 0.6384 - G Loss: 1.8397
Timestamp 20170805-20-15-10.964 -  Epoch 1/1 -  Step 480 -  D Loss: 0.5859 - G Loss: 2.5391
Timestamp 20170805-20-15-22.873 -  Epoch 1/1 -  Step 490 -  D Loss: 1.1367 - G Loss: 1.0803
Timestamp 20170805-20-15-34.783 -  Epoch 1/1 -  Step 500 -  D Loss: 0.9679 - G Loss: 1.0890

Timestamp 20170805-20-15-47.356 -  Epoch 1/1 -  Step 510 -  D Loss: 0.7404 - G Loss: 2.7417
Timestamp 20170805-20-15-59.256 -  Epoch 1/1 -  Step 520 -  D Loss: 0.9746 - G Loss: 1.9942
Timestamp 20170805-20-16-11.143 -  Epoch 1/1 -  Step 530 -  D Loss: 0.8230 - G Loss: 2.4660
Timestamp 20170805-20-16-23.030 -  Epoch 1/1 -  Step 540 -  D Loss: 0.8569 - G Loss: 1.5539
Timestamp 20170805-20-16-34.943 -  Epoch 1/1 -  Step 550 -  D Loss: 1.1575 - G Loss: 1.7777
Timestamp 20170805-20-16-46.828 -  Epoch 1/1 -  Step 560 -  D Loss: 0.9077 - G Loss: 1.3542
Timestamp 20170805-20-16-58.712 -  Epoch 1/1 -  Step 570 -  D Loss: 0.8227 - G Loss: 1.8265
Timestamp 20170805-20-17-10.621 -  Epoch 1/1 -  Step 580 -  D Loss: 1.6402 - G Loss: 0.6652
Timestamp 20170805-20-17-22.546 -  Epoch 1/1 -  Step 590 -  D Loss: 1.3822 - G Loss: 1.5119
Timestamp 20170805-20-17-34.440 -  Epoch 1/1 -  Step 600 -  D Loss: 1.2602 - G Loss: 1.0395

Timestamp 20170805-20-17-47.002 -  Epoch 1/1 -  Step 610 -  D Loss: 0.9136 - G Loss: 1.7258
Timestamp 20170805-20-17-58.879 -  Epoch 1/1 -  Step 620 -  D Loss: 1.0908 - G Loss: 1.2309
Timestamp 20170805-20-18-10.775 -  Epoch 1/1 -  Step 630 -  D Loss: 0.9182 - G Loss: 1.1273
Timestamp 20170805-20-18-22.676 -  Epoch 1/1 -  Step 640 -  D Loss: 0.8470 - G Loss: 1.6454
Timestamp 20170805-20-18-34.576 -  Epoch 1/1 -  Step 650 -  D Loss: 1.1015 - G Loss: 0.8767
Timestamp 20170805-20-18-46.470 -  Epoch 1/1 -  Step 660 -  D Loss: 1.4782 - G Loss: 1.0115
Timestamp 20170805-20-18-58.355 -  Epoch 1/1 -  Step 670 -  D Loss: 0.9653 - G Loss: 1.2330
Timestamp 20170805-20-19-10.256 -  Epoch 1/1 -  Step 680 -  D Loss: 1.0232 - G Loss: 1.1671
Timestamp 20170805-20-19-22.144 -  Epoch 1/1 -  Step 690 -  D Loss: 0.8116 - G Loss: 1.6586
Timestamp 20170805-20-19-34.047 -  Epoch 1/1 -  Step 700 -  D Loss: 1.3664 - G Loss: 0.8403

Timestamp 20170805-20-19-46.599 -  Epoch 1/1 -  Step 710 -  D Loss: 1.3560 - G Loss: 0.7816
Timestamp 20170805-20-19-58.488 -  Epoch 1/1 -  Step 720 -  D Loss: 0.9073 - G Loss: 1.6667
Timestamp 20170805-20-20-10.400 -  Epoch 1/1 -  Step 730 -  D Loss: 0.8680 - G Loss: 2.5225
Timestamp 20170805-20-20-22.281 -  Epoch 1/1 -  Step 740 -  D Loss: 1.5982 - G Loss: 0.8684
Timestamp 20170805-20-20-34.174 -  Epoch 1/1 -  Step 750 -  D Loss: 0.8479 - G Loss: 1.7806
Timestamp 20170805-20-20-46.090 -  Epoch 1/1 -  Step 760 -  D Loss: 0.7943 - G Loss: 1.7354
Timestamp 20170805-20-20-57.988 -  Epoch 1/1 -  Step 770 -  D Loss: 1.4838 - G Loss: 0.6000
Timestamp 20170805-20-21-09.873 -  Epoch 1/1 -  Step 780 -  D Loss: 1.7799 - G Loss: 0.5502
Timestamp 20170805-20-21-21.760 -  Epoch 1/1 -  Step 790 -  D Loss: 1.9848 - G Loss: 0.2312
Timestamp 20170805-20-21-33.664 -  Epoch 1/1 -  Step 800 -  D Loss: 1.5083 - G Loss: 0.6142

Timestamp 20170805-20-21-46.221 -  Epoch 1/1 -  Step 810 -  D Loss: 1.0728 - G Loss: 1.2273
Timestamp 20170805-20-21-58.117 -  Epoch 1/1 -  Step 820 -  D Loss: 1.1856 - G Loss: 1.1462
Timestamp 20170805-20-22-09.994 -  Epoch 1/1 -  Step 830 -  D Loss: 0.8615 - G Loss: 1.5874
Timestamp 20170805-20-22-21.877 -  Epoch 1/1 -  Step 840 -  D Loss: 0.5632 - G Loss: 2.2496
Timestamp 20170805-20-22-33.769 -  Epoch 1/1 -  Step 850 -  D Loss: 0.9135 - G Loss: 1.4230
Timestamp 20170805-20-22-45.660 -  Epoch 1/1 -  Step 860 -  D Loss: 1.1455 - G Loss: 0.8849
Timestamp 20170805-20-22-57.563 -  Epoch 1/1 -  Step 870 -  D Loss: 0.8219 - G Loss: 0.8537
Timestamp 20170805-20-23-09.461 -  Epoch 1/1 -  Step 880 -  D Loss: 0.8613 - G Loss: 1.4010
Timestamp 20170805-20-23-21.357 -  Epoch 1/1 -  Step 890 -  D Loss: 0.8916 - G Loss: 1.2641
Timestamp 20170805-20-23-33.261 -  Epoch 1/1 -  Step 900 -  D Loss: 0.9766 - G Loss: 0.8713

Timestamp 20170805-20-23-45.821 -  Epoch 1/1 -  Step 910 -  D Loss: 1.1104 - G Loss: 0.9317
Timestamp 20170805-20-23-57.722 -  Epoch 1/1 -  Step 920 -  D Loss: 2.0165 - G Loss: 1.7538
Timestamp 20170805-20-24-09.635 -  Epoch 1/1 -  Step 930 -  D Loss: 1.5257 - G Loss: 0.7163
Timestamp 20170805-20-24-21.533 -  Epoch 1/1 -  Step 940 -  D Loss: 0.9352 - G Loss: 0.9658
Timestamp 20170805-20-24-33.424 -  Epoch 1/1 -  Step 950 -  D Loss: 1.1749 - G Loss: 1.0307
Timestamp 20170805-20-24-45.311 -  Epoch 1/1 -  Step 960 -  D Loss: 1.1000 - G Loss: 1.4664
Timestamp 20170805-20-24-57.189 -  Epoch 1/1 -  Step 970 -  D Loss: 1.0843 - G Loss: 1.0299
Timestamp 20170805-20-25-09.063 -  Epoch 1/1 -  Step 980 -  D Loss: 1.2483 - G Loss: 1.5807
Timestamp 20170805-20-25-20.966 -  Epoch 1/1 -  Step 990 -  D Loss: 1.0352 - G Loss: 2.5007
Timestamp 20170805-20-25-32.857 -  Epoch 1/1 -  Step 1000 -  D Loss: 0.9717 - G Loss: 1.5381

batch_size = 16
z_dim = 100
learning_rate = 0.0002
beta1 = 0.5

tf.reset_default_graph()

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Timestamp 20170806-04-31-57.166 -  Epoch 1/1 -  Step 10 -  D Loss: 1.3363 - G Loss: 0.8602
Timestamp 20170806-04-32-13.808 -  Epoch 1/1 -  Step 20 -  D Loss: 0.8091 - G Loss: 3.2806
Timestamp 20170806-04-32-30.464 -  Epoch 1/1 -  Step 30 -  D Loss: 0.6637 - G Loss: 2.2792
Timestamp 20170806-04-32-47.128 -  Epoch 1/1 -  Step 40 -  D Loss: 0.5932 - G Loss: 3.1021
Timestamp 20170806-04-33-03.768 -  Epoch 1/1 -  Step 50 -  D Loss: 0.5628 - G Loss: 3.7488
Timestamp 20170806-04-33-20.416 -  Epoch 1/1 -  Step 60 -  D Loss: 0.9936 - G Loss: 1.8259
Timestamp 20170806-04-33-37.054 -  Epoch 1/1 -  Step 70 -  D Loss: 0.8542 - G Loss: 2.6979
Timestamp 20170806-04-33-53.697 -  Epoch 1/1 -  Step 80 -  D Loss: 0.9811 - G Loss: 3.2895
Timestamp 20170806-04-34-10.341 -  Epoch 1/1 -  Step 90 -  D Loss: 0.5536 - G Loss: 4.3284
Timestamp 20170806-04-34-26.986 -  Epoch 1/1 -  Step 100 -  D Loss: 0.6431 - G Loss: 2.6831

Timestamp 20170806-04-34-44.262 -  Epoch 1/1 -  Step 110 -  D Loss: 1.6780 - G Loss: 0.7400
Timestamp 20170806-04-35-00.894 -  Epoch 1/1 -  Step 120 -  D Loss: 0.6300 - G Loss: 3.6255
Timestamp 20170806-04-35-17.534 -  Epoch 1/1 -  Step 130 -  D Loss: 0.8542 - G Loss: 7.3327
Timestamp 20170806-04-35-34.207 -  Epoch 1/1 -  Step 140 -  D Loss: 0.4718 - G Loss: 4.5313
Timestamp 20170806-04-35-50.832 -  Epoch 1/1 -  Step 150 -  D Loss: 0.8467 - G Loss: 1.2472
Timestamp 20170806-04-36-07.457 -  Epoch 1/1 -  Step 160 -  D Loss: 1.9606 - G Loss: 0.6993
Timestamp 20170806-04-36-24.128 -  Epoch 1/1 -  Step 170 -  D Loss: 0.9050 - G Loss: 2.2913
Timestamp 20170806-04-36-40.765 -  Epoch 1/1 -  Step 180 -  D Loss: 0.7000 - G Loss: 2.1616
Timestamp 20170806-04-36-57.400 -  Epoch 1/1 -  Step 190 -  D Loss: 0.9409 - G Loss: 2.2140
Timestamp 20170806-04-37-14.016 -  Epoch 1/1 -  Step 200 -  D Loss: 0.8440 - G Loss: 2.0788

Timestamp 20170806-04-37-31.233 -  Epoch 1/1 -  Step 210 -  D Loss: 1.4659 - G Loss: 0.7342
Timestamp 20170806-04-37-47.847 -  Epoch 1/1 -  Step 220 -  D Loss: 0.9462 - G Loss: 2.5353
Timestamp 20170806-04-38-04.473 -  Epoch 1/1 -  Step 230 -  D Loss: 1.2387 - G Loss: 0.8576
Timestamp 20170806-04-38-21.089 -  Epoch 1/1 -  Step 240 -  D Loss: 1.6184 - G Loss: 0.9149
Timestamp 20170806-04-38-37.715 -  Epoch 1/1 -  Step 250 -  D Loss: 1.4301 - G Loss: 0.9769
Timestamp 20170806-04-38-54.344 -  Epoch 1/1 -  Step 260 -  D Loss: 1.1445 - G Loss: 1.2024
Timestamp 20170806-04-39-10.982 -  Epoch 1/1 -  Step 270 -  D Loss: 1.0722 - G Loss: 0.8610
Timestamp 20170806-04-39-27.589 -  Epoch 1/1 -  Step 280 -  D Loss: 0.9682 - G Loss: 1.5049
Timestamp 20170806-04-39-44.205 -  Epoch 1/1 -  Step 290 -  D Loss: 0.8318 - G Loss: 1.6806
Timestamp 20170806-04-40-00.818 -  Epoch 1/1 -  Step 300 -  D Loss: 0.9636 - G Loss: 1.7556

Timestamp 20170806-04-40-18.085 -  Epoch 1/1 -  Step 310 -  D Loss: 1.0702 - G Loss: 2.7176
Timestamp 20170806-04-40-34.705 -  Epoch 1/1 -  Step 320 -  D Loss: 1.3026 - G Loss: 1.3751
Timestamp 20170806-04-40-51.323 -  Epoch 1/1 -  Step 330 -  D Loss: 1.2347 - G Loss: 1.1099
Timestamp 20170806-04-41-07.954 -  Epoch 1/1 -  Step 340 -  D Loss: 1.1684 - G Loss: 0.9695
Timestamp 20170806-04-41-24.565 -  Epoch 1/1 -  Step 350 -  D Loss: 0.6319 - G Loss: 2.6704
Timestamp 20170806-04-41-41.189 -  Epoch 1/1 -  Step 360 -  D Loss: 1.0916 - G Loss: 2.7627
Timestamp 20170806-04-41-57.823 -  Epoch 1/1 -  Step 370 -  D Loss: 1.1327 - G Loss: 1.0298
Timestamp 20170806-04-42-14.437 -  Epoch 1/1 -  Step 380 -  D Loss: 1.1813 - G Loss: 0.9193
Timestamp 20170806-04-42-31.065 -  Epoch 1/1 -  Step 390 -  D Loss: 0.8111 - G Loss: 1.7377
Timestamp 20170806-04-42-47.691 -  Epoch 1/1 -  Step 400 -  D Loss: 0.8829 - G Loss: 1.2802

Timestamp 20170806-04-43-05.044 -  Epoch 1/1 -  Step 410 -  D Loss: 1.6494 - G Loss: 1.0181
Timestamp 20170806-04-43-21.701 -  Epoch 1/1 -  Step 420 -  D Loss: 0.9315 - G Loss: 1.0411
Timestamp 20170806-04-43-38.334 -  Epoch 1/1 -  Step 430 -  D Loss: 0.8723 - G Loss: 1.5805
Timestamp 20170806-04-43-54.964 -  Epoch 1/1 -  Step 440 -  D Loss: 1.0941 - G Loss: 1.3640
Timestamp 20170806-04-44-11.587 -  Epoch 1/1 -  Step 450 -  D Loss: 0.7728 - G Loss: 0.8816
Timestamp 20170806-04-44-28.212 -  Epoch 1/1 -  Step 460 -  D Loss: 0.6597 - G Loss: 1.9190
Timestamp 20170806-04-44-44.840 -  Epoch 1/1 -  Step 470 -  D Loss: 0.8259 - G Loss: 1.5367
Timestamp 20170806-04-45-01.466 -  Epoch 1/1 -  Step 480 -  D Loss: 0.5752 - G Loss: 2.4773
Timestamp 20170806-04-45-18.098 -  Epoch 1/1 -  Step 490 -  D Loss: 0.9869 - G Loss: 1.7823
Timestamp 20170806-04-45-34.735 -  Epoch 1/1 -  Step 500 -  D Loss: 0.7321 - G Loss: 1.4268

Timestamp 20170806-04-45-51.963 -  Epoch 1/1 -  Step 510 -  D Loss: 0.8912 - G Loss: 1.9894
Timestamp 20170806-04-46-08.603 -  Epoch 1/1 -  Step 520 -  D Loss: 1.4225 - G Loss: 0.5351
Timestamp 20170806-04-46-25.224 -  Epoch 1/1 -  Step 530 -  D Loss: 0.7015 - G Loss: 1.8034
Timestamp 20170806-04-46-41.863 -  Epoch 1/1 -  Step 540 -  D Loss: 0.5588 - G Loss: 1.9742
Timestamp 20170806-04-46-58.483 -  Epoch 1/1 -  Step 550 -  D Loss: 0.6100 - G Loss: 2.0397
Timestamp 20170806-04-47-15.094 -  Epoch 1/1 -  Step 560 -  D Loss: 0.9487 - G Loss: 2.4883
Timestamp 20170806-04-47-31.750 -  Epoch 1/1 -  Step 570 -  D Loss: 0.7618 - G Loss: 1.7543
Timestamp 20170806-04-47-48.364 -  Epoch 1/1 -  Step 580 -  D Loss: 1.4475 - G Loss: 2.8004
Timestamp 20170806-04-48-04.985 -  Epoch 1/1 -  Step 590 -  D Loss: 0.9168 - G Loss: 1.7664
Timestamp 20170806-04-48-21.622 -  Epoch 1/1 -  Step 600 -  D Loss: 0.9365 - G Loss: 1.2774

Timestamp 20170806-04-48-38.880 -  Epoch 1/1 -  Step 610 -  D Loss: 0.8372 - G Loss: 1.5094
Timestamp 20170806-04-48-55.513 -  Epoch 1/1 -  Step 620 -  D Loss: 1.3124 - G Loss: 0.5207
Timestamp 20170806-04-49-12.142 -  Epoch 1/1 -  Step 630 -  D Loss: 0.5129 - G Loss: 2.4255
Timestamp 20170806-04-49-28.769 -  Epoch 1/1 -  Step 640 -  D Loss: 1.1859 - G Loss: 2.5036
Timestamp 20170806-04-49-45.366 -  Epoch 1/1 -  Step 650 -  D Loss: 0.9392 - G Loss: 1.1803
Timestamp 20170806-04-50-01.988 -  Epoch 1/1 -  Step 660 -  D Loss: 0.7920 - G Loss: 1.7831
Timestamp 20170806-04-50-18.612 -  Epoch 1/1 -  Step 670 -  D Loss: 0.9904 - G Loss: 1.7999
Timestamp 20170806-04-50-35.234 -  Epoch 1/1 -  Step 680 -  D Loss: 2.0462 - G Loss: 0.4048
Timestamp 20170806-04-50-51.861 -  Epoch 1/1 -  Step 690 -  D Loss: 0.9307 - G Loss: 1.5385
Timestamp 20170806-04-51-08.491 -  Epoch 1/1 -  Step 700 -  D Loss: 1.0891 - G Loss: 1.7496

Timestamp 20170806-04-51-25.756 -  Epoch 1/1 -  Step 710 -  D Loss: 2.1666 - G Loss: 0.3856
Timestamp 20170806-04-51-42.409 -  Epoch 1/1 -  Step 720 -  D Loss: 0.3818 - G Loss: 4.5385
Timestamp 20170806-04-51-59.024 -  Epoch 1/1 -  Step 730 -  D Loss: 0.8134 - G Loss: 1.5436
Timestamp 20170806-04-52-15.658 -  Epoch 1/1 -  Step 740 -  D Loss: 0.5719 - G Loss: 1.9198
Timestamp 20170806-04-52-32.301 -  Epoch 1/1 -  Step 750 -  D Loss: 0.8087 - G Loss: 1.1429
Timestamp 20170806-04-52-48.916 -  Epoch 1/1 -  Step 760 -  D Loss: 1.1230 - G Loss: 1.9050
Timestamp 20170806-04-53-05.614 -  Epoch 1/1 -  Step 770 -  D Loss: 1.1254 - G Loss: 0.7880
Timestamp 20170806-04-53-22.340 -  Epoch 1/1 -  Step 780 -  D Loss: 1.2792 - G Loss: 1.7705
Timestamp 20170806-04-53-39.050 -  Epoch 1/1 -  Step 790 -  D Loss: 1.0407 - G Loss: 1.7626
Timestamp 20170806-04-53-55.742 -  Epoch 1/1 -  Step 800 -  D Loss: 0.8393 - G Loss: 1.4073

Timestamp 20170806-04-54-13.116 -  Epoch 1/1 -  Step 810 -  D Loss: 1.4924 - G Loss: 4.5066
Timestamp 20170806-04-54-29.837 -  Epoch 1/1 -  Step 820 -  D Loss: 0.7908 - G Loss: 1.6559
Timestamp 20170806-04-54-46.525 -  Epoch 1/1 -  Step 830 -  D Loss: 1.3191 - G Loss: 1.3561
Timestamp 20170806-04-55-03.208 -  Epoch 1/1 -  Step 840 -  D Loss: 0.7321 - G Loss: 2.2946
Timestamp 20170806-04-55-19.881 -  Epoch 1/1 -  Step 850 -  D Loss: 1.1127 - G Loss: 0.8722
Timestamp 20170806-04-55-36.529 -  Epoch 1/1 -  Step 860 -  D Loss: 0.6097 - G Loss: 2.3362
Timestamp 20170806-04-55-53.168 -  Epoch 1/1 -  Step 870 -  D Loss: 0.9553 - G Loss: 1.5836
Timestamp 20170806-04-56-09.808 -  Epoch 1/1 -  Step 880 -  D Loss: 1.0167 - G Loss: 1.1549
Timestamp 20170806-04-56-26.459 -  Epoch 1/1 -  Step 890 -  D Loss: 1.0836 - G Loss: 1.1680
Timestamp 20170806-04-56-43.077 -  Epoch 1/1 -  Step 900 -  D Loss: 1.1860 - G Loss: 0.7290

Timestamp 20170806-04-57-00.353 -  Epoch 1/1 -  Step 910 -  D Loss: 0.9428 - G Loss: 1.2215
Timestamp 20170806-04-57-16.986 -  Epoch 1/1 -  Step 920 -  D Loss: 0.7135 - G Loss: 1.7415
Timestamp 20170806-04-57-33.644 -  Epoch 1/1 -  Step 930 -  D Loss: 0.6923 - G Loss: 4.4217
Timestamp 20170806-04-57-50.290 -  Epoch 1/1 -  Step 940 -  D Loss: 1.1999 - G Loss: 1.7232
Timestamp 20170806-04-58-06.951 -  Epoch 1/1 -  Step 950 -  D Loss: 0.9554 - G Loss: 2.6222
Timestamp 20170806-04-58-23.605 -  Epoch 1/1 -  Step 960 -  D Loss: 1.0688 - G Loss: 1.1556
Timestamp 20170806-04-58-40.256 -  Epoch 1/1 -  Step 970 -  D Loss: 0.5570 - G Loss: 1.9077
Timestamp 20170806-04-58-56.886 -  Epoch 1/1 -  Step 980 -  D Loss: 0.9809 - G Loss: 1.3348
Timestamp 20170806-04-59-13.527 -  Epoch 1/1 -  Step 990 -  D Loss: 1.5010 - G Loss: 0.4302
Timestamp 20170806-04-59-30.172 -  Epoch 1/1 -  Step 1000 -  D Loss: 0.9442 - G Loss: 0.9294

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.